Spark
Sign In
MCP Connector

Debug AI Applications Remotely with GDB

Drives GDB remotely via GDB/MI - breakpoints, stepping, stack/register/memory inspection - with support for concurrent multi-session debugging.

Connect
Works with gdbmcp
Lix Zhou
Own this? Claim it

Maintainer of this project? Claim this page to edit the listing.

86
Spark score
out of 100
Updated 10 months ago
Fresher alternatives ↓
Version 0.2.3
Models
universal
Add to Favorites

Why it matters

Enable remote debugging of AI assistants by providing a GDB/MI protocol server. Facilitates breakpoint management, execution control, and multi-session debugging for enhanced development and troubleshooting.

Outcomes

What it gets done

01

Create and manage GDB debugging sessions remotely.

02

Control program execution including start, stop, step, and continue.

03

Set, delete, and retrieve breakpoints for targeted debugging.

04

Inspect stack frames, local variables, and registers during debugging.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/vb-gdb | bash

Capabilities

Tools your agent gets

create_session

Create a new GDB debugging session

get_session

Get information about a specific session

get_all_sessions

Get all sessions

close_session

Close a session

start_debugging

Start debugging

stop_debugging

Stop debugging

continue_execution

Continue execution

step_execution

Step to the next line

+7 tools

Overview

GDB MCP Server

This MCP server implements GDB/MI to give an AI assistant remote debugging capabilities - breakpoints, execution control, and stack/register/memory inspection - with concurrent multi-session support. Use it when an assistant needs to actively debug a running program rather than analyze code statically. Requires GDB installed on the system.

What it does

MCP Server GDB implements the GDB/MI protocol as an MCP server, giving AI assistants remote application debugging capabilities: creating and managing GDB debug sessions, setting and managing breakpoints, controlling program execution (run, pause, step into, step over), and inspecting stack frames, local variables, registers, and raw memory. It supports concurrent multi-session debugging, so multiple debug sessions can run side by side, and includes a built-in TUI (work in progress) for inspecting how the AI agent is behaving during a debugging session, useful for refining prompts.

When to use - and when NOT to

Use this when you want an AI assistant to debug a running program directly - setting breakpoints, stepping through execution, examining the call stack, local variables, or memory contents at a specific point - rather than reasoning about code statically. It requires GDB itself to be available on the system since this server drives GDB through the machine-oriented GDB/MI interface. It supports stdio (default) or SSE transport (default http://127.0.0.1:8080).

Capabilities

Session: create_session, get_session, get_all_sessions, close_session. Debug control: start_debugging, stop_debugging, continue_execution, step_execution (step into), next_execution (step over). Breakpoints: get_breakpoints, set_breakpoint, delete_breakpoint. Debug info: get_stack_frames, get_local_variables, get_registers, read_memory.

How to install

Download a pre-built binary for your platform from the releases page, or build from source:

cargo build --release
cargo run

Or with Nix: nix run . (local) or nix run "git+https://github.com/pansila/mcp_server_gdb.git" (remote). Run directly with ./mcp-server-gdb. Server IP, port, and GDB command timeout are configurable via src/config.rs or environment variables.

Who it's for

Developers who want an AI assistant to actively debug a running program via GDB - setting breakpoints, stepping through code, and inspecting live state - rather than just reading source statically, including workflows that debug multiple sessions concurrently. Licensed MIT.

Source README

MCP Server GDB

A GDB/MI protocol server based on the MCP protocol, providing remote application debugging capabilities with AI assistants.

Features

  • Create and manage GDB debug sessions
  • Set and manage breakpoints
  • View stack information and variables
  • Control program execution (run, pause, step, etc.)
  • Support concurrent multi-session debugging
  • A built-in TUI to inspect agent behaviors so that you can improve your prompt (WIP)

Installation

Pre-built Binaries

Find the binaries in the release page, choose one per your working platform, then you can run it directly.

Build From Source

Clone the repository and build it by cargo

cargo build --release
cargo run

Using Nix

If you have Nix installed, you can run the project without cloning:

Run locally (after cloning)
nix run .
Run remotely from GitHub
nix run "git+https://github.com/pansila/mcp_server_gdb.git" -- --help
Development environment

To enter a development shell with all dependencies:

nix develop

Usage

  1. Just run it directly: ./mcp-server-gdb
  2. The server supports two transport modes:
    • Stdio (default): Standard input/output transport
    • SSE: Server-Sent Events transport, default at http://127.0.0.1:8080

Configuration

You can adjust server configuration by modifying the src/config.rs file or by environment variables:

  • Server IP Address
  • Server port
  • GDB command timeout time (in seconds)

Supported MCP Tools

Session Management

  • create_session - Create a new GDB debugging session
  • get_session - Get specific session information
  • get_all_sessions - Get all sessions
  • close_session - Close session

Debug Control

  • start_debugging - Start debugging
  • stop_debugging - Stop debugging
  • continue_execution - Continue execution
  • step_execution - Step into next line
  • next_execution - Step over next line

Breakpoint Management

  • get_breakpoints - Get breakpoint list
  • set_breakpoint - Set breakpoint
  • delete_breakpoint - Delete breakpoint

Debug Information

  • get_stack_frames - Get stack frame information
  • get_local_variables - Get local variables
  • get_registers - Get registers
  • read_memory - Read memory contents

FAQ

Common questions

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.